Some parts of my previous message are premature---I haven't actually tried exactly this code on x86, and I should also try David Rientjes's patch. But I'd still like to know how this has been tested on ARM. Thanks! On Fri, Nov 2, 2012 at 12:59 PM, Luigi Semenzato <semenzato@xxxxxxxxxx> wrote: > Does anybody have any information on the status of zram on ARM? > Specifically, how much it has been tested. > > I noticed that zram and zsmalloc on ToT no longer have the x86 > restriction, and they compile fine on our 3.4 branch. Sadly, that's > where my luck ends. > > When I run my standard Chrome load (which just opens a bunch of > memory-intensive browser tabs), Chrome dies shortly after the system > starts swapping pages out. For instance, here's are the SI and SO > fields of "vmstat 1": > > si so > 0 0 > 0 0 > 0 0 > 0 0 > 0 0 > 0 0 > 0 0 > 0 0 > 0 168 > 0 0 > 0 924 > 188 26332 > 520 30672 > 1304 32208 > 2360 30804 > 18836 24832 > <--- chrome dies here > 6496 0 > 892 0 > 260 0 > 8 0 > > I also have a simpler load: a program that allocates memory non-stop, > and fills part of it with data from /dev/urandom (to simulate the > observed compressibility). The program never reads its data though, so > it doesn't get swapped back in, as in the previous load. This runs > for a while and partially fills the swap device, then the system > hangs. > > Deja vu, eh? I am running this with my patch, which may result in > extra OOM kills. Interestingly, a few threads are blocked in > exit_mm(), but not on a page fault. Most processes are in > congestion_wait(), so this is probably not the same situation I was > seeing earlier. > > Anyway, I am attaching the output of SysRQ-X with lots of stack > traces. Thank you very much for any information! > > Luigi -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>